Research on flexible job shop scheduling problems based on deep reinforcement learning

Abstract The flexible job shop scheduling problem is a typical class of combinatorial optimization problems and one of the classical NP-hard problems. By optimizing the scheduling scheme, it can realize the reasonable arrangement of the production line, improve the utilization rate of the equipment, reduce the production cycle, and thus enhance the competitiveness of the enterprise. Therefore, this article focuses on the complexity of solving flexible job shop scheduling problems, with the optimization objective of minimizing the maximum completion time. A dual channel state feature is established as the input of the neural network, and 22 combined heuristic scheduling rules are used as the output of the neural network. Action effectiveness exploration and utilization strategies are adopted to select the best action, and a reward function based on machine utilization is designed. Finally, standard cases were used to experiment with the algorithm’s solving time, verifying the effectiveness and generalization of the proposed algorithm.